HAL Id: tel-02864456
https://tel.archives-ouvertes.fr/tel-02864456
Submitted on 11 Jun 2020HAL is a multi-disciplinary open access
archive for the deposit and dissemination of sci-entific research documents, whether they are pub-lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.
sensors networks
Taha Al-Wajeeh
To cite this version:
Taha Al-Wajeeh. Efficient radio channel modeling for urban wireless sensors networks. Electromag-netism. Université de Poitiers, 2018. English. �NNT : 2018POIT2314�. �tel-02864456�
U n i v e r s i t ´
e
d e
P o i t i e r s
— Sciences Fondamentales et Appliqu´ees —
TH`
ESE
pour l’obtention du Grade de
DOCTEUR DE L’UNIVERSIT ´E DE POITIERS (Facult´e des Sciences Fondamentales et Appliqu´ees)
(Diplˆome National - Arrˆet´e du 25 Avril 2002)
´
Ecole Doctorale : Sciences et Ing´enierie pour l’Information, Math´ematiques (S2IM) Secteur de Recherche : ´Electronique, micro´electronique, nano´electronique et micro-ondes
pr´esent´ee par
ALWAJEEH Taha
Mod´
elisation efficace du canal radio pour
les r´
eseaux de capteurs urbains.
.
Efficient radio channel modeling for urban
wireless sensor networks.
Soutenue le 13 D´ecembre 2018 devant la Commission d’Examen compos´ee de :
M. UGUEN Bernard, Professeur, Universit´e de Rennes 1 . . . Rapporteur M. KHENCHAF Ali, Professeur, ENSTA Bretagne . . . Rapporteur M. RAOOF Kosa¨ı, Professeur, Universit´e du Maine . . . Examinateur M. BERNARD Lo¨ıc, Docteur, HDR, Institut de recherche de Saint-Louis . . . Examinateur M. MARC Olivier, Ing´enieur, Virtualys, Brest . . . Examinateur M. VAUZELLE Rodolphe, Professeur, Universit´e de Poitiers . . . Directeur de Th`ese M. COMBEAU Pierre, Maˆıtre de Conf´erences, Universit´e de Poitiers . . . Co-encadrant de Th`ese M. BOUNCEUR Ahc`ene, Maˆıtre de Conf´erences, Universit´e de Bretagne Occidentale Co-encadrant de Th`ese
...God blessed me with a ...wonderful daughter
Acknowledgments
First and foremost, I would like to express my deepest gratitude and appreciation to Mr. Rodolphe VAUZELLE, Mr. Pierre Combeau, and Mr. Ahc`ene Bounceur for their invaluable guidance, immense knowledge, constant support, consistent encouragement, and patient ad-vice throughout this journey. Thank you all for your unwavering support at all times. A special gratitude goes to Pierre Combeau who, besides his academic support, shared my per-sonal concerns throughout my dissertation.
I also wish to sincerely thank Mr. Bernard UGUEN, Professor at the University of Rennes 1, and Mr. Ali KHENCHAF, Professor at ENSTA Bretagne, for their interest in this work by accepting to be reporters on this thesis.
I would like to thank Mr. Kosa¨ı RAOOF, Professor at the University of Maine, Mr. Lo¨ıc BERNARD, Doctor, HDR, at the research institute of Saint-Louis, and Mr. Olivier MARC, Engineer at Virtualys, for the honor that has been conferred on me by agreeing to chair the thesis committee.
On a personal level, I would like to express my profound gratitude to my GREAT parents, my lovely wife, my supportive siblings, my little princess, who provided me with moral and emotional support all along the way, and I undoubtedly could not have done this accomplish-ment without them.
Finally, I conclude by expressing my thanks to my friends and to every single member of XLIM research institute of the University of Poitiers for their hospitality, collaboration, assistance, and the fruitful discussions.
Table of contents
General introduction 13
1 Context and motivation 15
1.1 Introduction . . . 16
1.2 Smart cities . . . 16
1.3 Wireless sensor networks . . . 18
1.4 WSN protocols . . . 19
1.4.1 Short range communication protocols . . . 20
1.4.1.1 IEEE Std 802.15.4 . . . 20
1.4.1.2 IEEE 802.15.4 derived standards . . . 20
1.4.2 Low-Power Wide-Area Network (LPWAN) protocols . . . 22
1.4.2.1 Sigfox . . . 22
1.4.2.2 LoRa . . . 23
1.5 Simulation tools for wireless sensor networks . . . 23
1.6 PERSEPTEUR project . . . 26
1.6.1 Project objectives . . . 26
1.6.2 Project parteners . . . 27
1.7 Thesis motivation and objectives . . . 28
1.8 Thesis Contributions . . . 30
1.9 Conclusion . . . 30
2 Background and state of the art 31 2.1 Introduction . . . 33
2.2 Radio Channel fundamentals . . . 33
2.2.1 Multipath propagation . . . 33
2.2.2 Different Scales of Attenuation . . . 36
2.3 Parameters of Multipath Radio Channels . . . 37
2.3.1 Time dispersion parameters . . . 37
2.3.2 Coherence bandwidth . . . 38
2.3.3 Doppler spread and coherence time . . . 38
2.3.5 Macro, Micro, Pico, and Femto Cells . . . 40
2.4 Review of empirical propagation radio models for WSNs . . . 41
2.4.1 Free Space Model . . . 41
2.4.2 Adapted Free Space Model . . . 42
2.4.3 Simplified Two-Ray Ground Reflection Model . . . 42
2.4.4 Free Space and Simplified Two-Ray Hybrid Model . . . 44
2.4.5 Two-Ray Ground Reflection Model . . . 44
2.4.6 Free Outdoor Model (FOM) . . . 45
2.4.7 Log Normal Model Shadowing Model . . . 46
2.4.8 Survey of radio propagation models in WSN Simulators . . . 48
2.5 Deterministic modeling . . . 50
2.5.1 Maxwell’s Equations . . . 50
2.5.2 Electrical properties of propagation medium . . . 51
2.5.3 Rigorous solutions to Maxwell’s equations . . . 52
2.5.4 Helmholtz wave Equation . . . 52
2.5.5 High frequency asymptotic methods . . . 52
2.5.5.1 Geometric Optics and its extensions . . . 53
2.5.5.2 Wave types . . . 54
2.5.5.3 Reflection . . . 55
2.5.5.4 Diffraction . . . 57
2.5.5.5 The geometrical theory of diffraction . . . 58
2.5.5.6 The Uniform theory of diffraction . . . 59
2.6 Ray Tracing Techniques for Radio Propagation . . . 61
2.6.1 Ray-launching method . . . 61
2.6.2 Image method . . . 62
2.6.3 Hybrid method . . . 63
2.6.4 Other methods . . . 63
2.7 Ray-Tracing Acceleration Techniques . . . 63
2.7.1 Dimension Reduction . . . 63
2.7.2 Space Division . . . 64
2.7.3 Visibility Graphs . . . 64
2.7.4 GPU Acceleration . . . 64
2.8 Formulation of the research problem . . . 64
2.9 Proposed Solution . . . 64
2.10 Conclusion . . . 67
3 Visibility tree and acceleration techniques 69 3.1 Introduction . . . 70
3.2 Input Data Bases . . . 70
3.2.1 Description of the propagation environment . . . 70
3.2.2 Description of test scenes . . . 71
3.3 Ray-tracing based on the visibility tree . . . 72
3.3.1 Creation of a horizontal 2D visibility tree . . . 73
3.3.1.1 Principle and data structure . . . 73
3.3.1.2 Horizontal plane profile from a 3D city map . . . 74
3.3.1.3 Initialization of the visible zones . . . 74
5
3.3.1.5 Visible zones . . . 78
3.3.1.6 Reflected zones . . . 79
3.3.1.7 Diffracted zones . . . 80
3.3.2 Calculating propagation paths from a visibility tree . . . 81
3.3.2.1 Propagation paths in the horizontal plane . . . 81
3.3.2.2 Propagation paths in 3D . . . 84
3.3.2.3 Ground reflected path . . . 85
3.3.2.4 Electric field Computation . . . 85
3.4 Model validation . . . 85
3.4.1 Punctual validation with a standard 3D full ray-tracing tool . . . 85
3.4.1.1 Reflection . . . 86
3.4.1.2 Diffraction . . . 88
3.4.2 Model validation with field measurements . . . 89
3.4.2.1 Scene modeling . . . 89
3.4.2.2 Antenna radiation pattern . . . 90
3.4.2.3 Model performance without power averaging . . . 91
3.4.2.4 Time gain . . . 92
3.4.2.5 Local average power estimation of mobile radio signals . . . . 93
3.4.2.6 Model performance with power averaging . . . 93
3.4.2.7 Sources of error . . . 95
3.4.3 Narrow-band and Wide-band simulations . . . 95
3.5 Acceleration techniques . . . 97
3.5.1 Scene preparation acceleration algorithms . . . 97
3.5.2 Maximum number of propagation paths . . . 98
3.5.2.1 Hypothesis . . . 98
3.5.2.2 Description of the algorithm . . . 98
3.5.2.3 Accuracy and Time gain . . . 98
3.5.3 Limited area . . . 99
3.5.3.1 Hypothesis . . . 99
3.5.3.2 Description of the algorithm . . . 99
3.5.3.3 Accuracy and Time gain . . . 100
3.5.4 Visibility trees storage (Pre-processing) . . . 104
3.5.4.1 Hypothesis . . . 105
3.5.4.2 Description of the algorithm . . . 105
3.5.4.3 Time gain and data size . . . 108
3.5.4.4 Accuracy . . . 108
3.6 Optimal number of interactions . . . 109
3.6.1 Test scenarios . . . 109
3.6.2 Parametric study . . . 110
3.7 Conclusion . . . 111
4 Propagation in the vertical plane 113 4.1 Introduction . . . 114
4.2 Downtown Munich map and measurements . . . 115
4.2.1 Measurement path Metro 200 . . . 116
4.2.2 Measurement path Metro 201 . . . 116
4.3 Problem statement . . . 118
4.4 Vertical plane extraction . . . 120
4.5 Vertical Propagation Paths . . . 121
4.6 State of the art review . . . 122
4.6.1 Multiple knife-edge diffraction . . . 123
4.6.2 Multiple edge Diffraction Integral . . . 123
4.6.3 GTD/UTD . . . 124
4.6.4 UTD Slope diffraction . . . 124
4.6.5 UTD-slope diffraction with distance parameter forcing . . . 125
4.6.5.1 Principle . . . 125
4.6.5.2 Implementation . . . 126
4.6.5.3 Results and limitations . . . 127
4.6.6 Capolino and Albani Method . . . 129
4.6.6.1 Geometry and double diffraction coefficient . . . 129
4.6.7 Application to 2D configurations . . . 131
4.6.8 3D Double diffraction coefficient . . . 134
4.6.9 Application to 3D configurations . . . 135
4.7 Model Validation . . . 137
4.7.1 Global validation via Munich measurements . . . 137
4.7.1.1 Local average power estimation . . . 137
4.7.1.2 Model performance . . . 138
4.7.1.3 Sources of error . . . 139
4.7.1.4 Curve smoothing . . . 139
4.7.2 Computational time . . . 141
4.8 Vertical propagation impact on Charles de Gaulle - ´Etoile scene . . . 141
4.9 Conclusion . . . 142
5 Integration and case study 143 5.1 Introduction . . . 144
5.2 Integration into CupCarbon . . . 144
5.2.1 CupCarbon . . . 144
5.2.2 Integration procedure . . . 145
5.2.2.1 Application Programming Interface . . . 145
5.3 Geometry databases . . . 146
5.3.1 GPS coordinate system conversion . . . 146
5.3.2 Geometry database simplification . . . 146
5.3.3 Precision of geometry databases . . . 149
5.4 API validation . . . 149
5.5 Mobility case study . . . 152
5.6 Conclusion . . . 158
List of figures
1.1 Smart city applications. Image source : Trends in smart city development : Case studies
and recommendations [DS16]. . . 17
1.2 Block diagram of a sensor node. . . 18
1.3 Classification of WSN protocols. . . 19
1.4 Simulator structure. . . 28
1.5 Accuracy computational time trade-off. . . 30
2.1 Urban propagation scenario. . . 34
2.2 Bello functions. . . 36
2.3 Path loss, shadowing and fast fading. . . 37
2.4 Macro, Micro, Pico, and Femto Cells. . . 40
2.5 Free space model vs. adapted model. . . 42
2.6 Two-ray ground reflection model. . . 43
2.7 Simplified two-ray ground reflection Model. . . 43
2.8 Simplified two-ray, two-ray, and free space models. . . 45
2.9 Free outdoor model. . . 46
2.10 Distance-height relationship. . . 46
2.11 Simulation results for the two-slope log normal model. . . 48
2.12 Astigmatic ray tube. . . 54
2.13 Wavefronts : a. spherical wave, b. cylindrical wave, c. plane wave. . . 55
2.14 Incident, reflected, and transmitted wave at a smooth plane.. . . 56
2.15 The cone of diffracted rays.. . . 58
2.16 Geometry for diffraction by a wedge. . . 59
2.17 Transition Function. . . 60
2.18 Ray-Launching method. . . 62
2.19 The image method.. . . 62
2.20 Proposed solution in the horizontal plane. . . 66
3.1 Considered urban environments. . . 72
3.2 Tree structure. . . 73
3.4 Initial visible zones. . . 75
3.5 Line discretization. . . 76
3.6 Zone discretization.. . . 77
3.7 Discretization direction. . . 77
3.8 Visible zones. . . 78
3.9 Visibility tree - layer 1 : visible zones. . . 78
3.10 Reflected zones. . . 79
3.11 Visibility tree - with reflection. . . 79
3.12 Keller cone in 2D. . . 80
3.13 Diffracted zones. . . 80
3.14 Visibility tree - with reflection and diffraction at the point P .. . . 81
3.15 Diffracted zones. . . 82
3.16 Visibility tree - nodes including the receiver . . . 82
3.17 Zones including the receiver Rx. . . 83
3.18 Grid resolution impact.. . . 84
3.19 Paths transformation into 3D. . . 84
3.20 Paths transformation into 3D. . . 85
3.21 Punctual model validation - Reflection. . . 86
3.22 Model validation - Diffraction. . . 88
3.23 Charles de Gaulle - ´Etoile, Paris.. . . 90
3.24 Radiation pattern of a dipole. . . 91
3.25 Global model validation. . . 92
3.26 Model validation - local average power estimation. . . 94
3.27 Major source of error - ´Etoile, Paris. . . 95
3.28 Coverage prediction. . . 96
3.29 RMS delay spread. . . 96
3.30 Diffraction Points. . . 97
3.31 Limited zone algorithm . . . 100
3.32 Limited area test scenarios. . . 103
3.33 Simple example of a visibility tree - 2R1D. . . 105
3.34 Search for saved visibility tree. . . 107
3.35 Test scenarios - optimal number of interactions. . . 109
3.36 Guide for the optimum number of interactions - (150 m, grid of receivers). . . 110
4.1 Munich3D. . . 115
4.2 Munich site - Metro 200 route. . . 116
4.3 Munich site - Metro 201 route. . . 117
4.4 Munich site - Metro 202 route. . . 118
4.5 Munich site - Metro 202 route : propagation in horizontal plane . . . 119
4.6 Google earth view. . . 119
4.7 Vertical plane extraction.. . . 120
4.8 Munich city - vertical plane extraction. . . 121
4.9 Vertical propagation paths - implemented options. . . 122
4.10 Geometry for Holm expression.. . . 125
4.11 Propagation path over multiple wedges with unequal heights. . . 126
4.12 Validation of the distance parameter forcing method. . . 128
9
4.14 Generalized Fresnel Integrals validation. . . 131
4.15 Geometry for double diffraction by a flat plate. . . 131
4.16 Capolino-Albani vs. UTD coefficients in 2D configurations. . . 132
4.17 Continuity of Capolino-Albani diffraction coefficient.. . . 133
4.18 Geometry for double diffraction by a plate - different plate widths. . . 133
4.19 Simulation results of different thicknesses. . . 134
4.20 3D Geometry for Capolino-Albani method. . . 135
4.21 Capolino and Albani 3D test configuration. . . 135
4.22 Capolino and Albani 3D coefficient validation. . . 136
4.23 Averaging sector - Munich site.. . . 137
4.24 Model performance. . . 138
4.25 Model performance - curve smoothing. . . 140
4.26 Simulation results of Charles de Gaulle - ´Etoile, scene with/without the vertical plane. 142 5.1 CupCarbon user interface. . . 145
5.2 Geometry simplification. . . 148
5.3 API validation - CupCarbon user interface. . . 149
5.4 Charles de Gaulle - ´Etoile, Paris.. . . 150
5.5 Epsilon effect on accuracy. . . 151
5.6 Mobility scenario. . . 152
5.7 Attenuation values for S10at different time instants.. . . 154
5.8 Deterministic attenuation in function of distance. . . 155
5.9 Deterministic vs. Log-Normal model.. . . 156
5.10 Package error rate. . . 157
List of tables
1.1 IEEE 802.15.4 derived standards . . . 21
1.2 LoRaWAN specifications . . . 23
2.2 Fading types . . . 39
2.3 Some typical values of path loss exponent and shadowing deviation . . . 47
2.4 Two-slope log-normal model parameters . . . 47
2.5 Propagation models implemented in some WSN simulators. . . 50
3.1 Model validation - Reflection without ground reflected ray. . . 87
3.2 Model validation - Reflection with ground reflected ray. . . 87
3.3 Model validation - Diffraction in LOS configuration. . . 88
3.4 Model validation - Diffraction in NLOS configuration . . . 89
3.5 Comparison of computational time. . . 93
3.6 Model performance - error evaluation. . . 94
3.7 Accuracy and gain in time - limited number of propagation paths.. . . 99
3.8 Accuracy vs. gain in time. . . 101
3.9 Accuracy vs. gain in time. . . 102
3.10 Gain in time. . . 104
3.11 Data structure for pre-calculated trees - integer table. . . 106
3.12 EXECUTION TIME AND SIZE FOR PRE-CALCULATED VISIBILITY TREES . . 108
3.13 STEP SIZE VS MEAN ERROR.. . . 108
4.1 Test parameters . . . 132
4.2 Test parameters . . . 134
4.3 Error between the measurements and model prediction (HP : 3R1D, VP : Capolino and Albani Method) . . . 139
4.4 Error between the measurements and model prediction (HP :3R1D, VP : classical UTD) . . . 139
4.5 Error between the measurements and model prediction (HP : 3R1D, VP : Capolino and Albani Method) with curve smoothing. . . 140
5.1 Geometry simplification. . . 147
5.2 API validation - error estimation for different values of . . . 150
5.3 PER. . . 157
General introduction
This thesis is a part of a national research project funded by the French National Research Agency (ANR) and supported by the National Center for Scientific Research (CNRS). The work presented in this thesis was carried out at XLIM research institute of the University of Poitiers under the direction of Mr. Rodolphe Vauzelle, Professor at the University of Poitier, Mr. Pierre Combeau, Senior Lecturer at the University of Poitiers, and Mr. Ahc`ene Bounceur Associate Professor at the University of Brest (UBO).
Nowadays, urban services are presented in a different way because new approaches of ma-naging cities are replacing the conventional ones. In fact, the vision of mama-naging cities was reshaped due to the wirelessly connected elements offered by new technologies. These new technologies improved the quality and efficiency of urban services for a wide range of appli-cations in various domains. Wireless Sensor Networks and Internet of Things are the leading technologies that introduced the concept of “Smart City”.
The radio channel is an unavoidable element in any wireless systems. Signals passing via the radio channel are subjected to impairments introduced by the channel. Modeling the radio channel for the wirelessly connected nodes within the wireless network simulators is the key point of the here presented thesis. The objective of this thesis is therefore to develop radio methods for modeling electromagnetic waves for outdoor urban environments. These models should guarantee a high degree of precision in a city-wide scale under tight time constraints. The models should also expect a large number of connected nodes.
The thesis consists of five chapters. The first chapter presents the general context, in which this thesis was carried out. It also discusses : the concept of the “Smart Cities” and their application domains, Wireless Sensor Networks, Internet of Things, and wireless net-work simulators and protocols. It also outlines the required specifications and the expected outcome of this work.
Chapter 2 provides a detailed summary of the relevant literature of : the radio channel fundamentals, radio wave interactions with the environment, channel modeling approaches, and the advantages and disadvantages of each method. Accordingly, it formulates the research problem, and then it concludes with a proposal for answering the research problem in an ef-ficient way.
Chapter 3 will focus on developing a radio model for the configurations where the predo-minant propagation mechanism is occurring in the horizontal plane i.e the antennas height is lower than the average height of the surrounding environment. The adopted approach will be based on the visibility technique between the transmitter and the obstacles in the propa-gation environment. It will also be based on new acceleration techniques in order to ensure the rapidity and accuracy of the model. Results will be evaluated in terms of precision and execution time.
Chapter 4 starts by evaluating the validity of the radio model of chapter 3 when the an-tennas height is higher than the surroundings. Consequently, the objective of chapter 4 is to extend the existing model to be valid for all urban configurations. Therefore, it will integrate an appropriate radio model in the vertical plane. Simulation results will be compared with measurements conducted within a European project.
Chapter 5 will focus on the integration of the radio models into the final platform (i.e wi-reless sensor simulator). It first starts by introducing the platform to which the radio models are going to be integrated. It also details the integration process, and then it lists the encoun-tered problems during this phase, especially due to the geometry databases. It also shows how these issues will be resolved. Chapter 5 concludes with a real case study of a number of sensors in an urban configuration. The study evaluates the impact on of integrating accurate radio channel to the wireless sensor simulator.
Chapter 1
Context and motivation
Contents
1.1 Introduction . . . 16
1.2 Smart cities . . . 16
1.3 Wireless sensor networks . . . 18
1.4 WSN protocols . . . 19
1.4.1 Short range communication protocols . . . 20
1.4.2 Low-Power Wide-Area Network (LPWAN) protocols . . . 22
1.5 Simulation tools for wireless sensor networks . . . 23
1.6 PERSEPTEUR project . . . 26
1.6.1 Project objectives . . . 26
1.6.2 Project parteners . . . 27
1.7 Thesis motivation and objectives . . . 28
1.8 Thesis Contributions . . . 30
1.1
Introduction
This introductory chapter intends to present the context, objectives, and expected out-comes for the here presented thesis. In this regard, it will be important to start by introducing a set of general concepts and definitions that will seen throughout this chapter.
The discussion in chapter 1 attempts to address three main aspects. To that end, this chapter consists of three parts : the first part introduces the concept of the “Smart Cities”, wireless sensor networks (WSNs), and wireless sensor network simulators. Then, the second part intends to present the research project “PERSEPTEUR”, in which this thesis was carried out. Furthermore, this part presents the project partners and a brief description of their main roles in the project. Finally, part three will focus on the thesis itself, specifically, on objectives and contributions.
1.2
Smart cities
It was estimated in 2014 that 54% of the world population lived in urban areas, roughly 3.3 billion people. This percentage is expected to increase to 66% by 2030, or roughly to 5 bil-lion people[Lea17]. This massive increase leads to new avenues of research for new approaches to manage cities and to offer urban services in a different way. Moreover, the increase of in-terconnected elements in cities’ infrastructure due to new technologies has also reshaped the vision of managing cities[Ers17]. This, in turn, leads to the concept of “Smart City”.
There are many angles from which smart cities can be defined. Generally speaking, a smart city refers to a city which uses new technologies and innovative ideas to improve the quality and efficiency of urban services in order to improve the lives of people. New solutions and new approaches are intended to improve public transportation, city traffic, energy resources, water management, environmental protection, surveillance, urban services, etc.
Technically speaking, smart cities involve three elements : information and communica-tion technologies (ICTs) that collect data ; analytical tools that convert the collected data into useful information ; and an application to analyze that real-time information.
Smart city concept involves a wide range of applications in various domains. A simple practical example of these applications is the smart parking project, called “Connected Par-king”, carried out with Libelium World in Montpellier city to quickly find a parking, reducing considerably searching time for a free parking space [Lib]. A recent research published by the American University with the National League of Cities [DS16] presents graphically a number of applications for smart cities as shown in figure 1.1, which lists some interesting applications such as :
1. Smart transportation systems. 2. Water monitoring.
3. Smart parking. 4. Bridge systems.
1.2 Smart cities 17 5. Autonomous cars. 6. Waste management. 7. Lighting management. 8. Fire detection. 9. Energy management. 10. Solar panels monitoring.
11. Smart freight/tracking systems. 12. Vehicle fleet communication. 13. Drone applications.
14. Surveillance camera. 15. Body cameras.
16. Wearable detection sensors so that people can be a part of system.
Figure 1.1 –Smart city applications. Image source : Trends in smart city development : Case studies and recommendations [DS16].
Currently, smart city solutions are mainly based on communication technologies such as Wireless Sensor Networks (WSN) and the Internet of Things (IoT) [GMM+17]. However, in the very near future, smart city solutions will rely on 5G networks. In fact, 5G networks have a new vision. 5G networks intend to target every single element of future life because 5G networks define three service categories : enhanced mobile broadband (eMBB), massive machine-type communication (mMTC), and ultra-reliable and low latency communication (uRLLC) [XZS16]. Anyway, the discussion here will be limited to the current vision of smart cities.
1.3
Wireless sensor networks
Recently, several sensor-based technologies are being widely used in a wide range of do-mains, particularly in monitoring and surveillance. A wireless sensor network is simply a deployment of a large number of autonomous and self-configured devices equipped with sen-sors to monitor a number of different phenomena of interest. These devices, which are called sensor nodes, are typically tiny, low-cost, and battery-operated devices. The data collected by these sensors are collaboratively passed through the network to a central point (i.e. a base station) that gathers data from all sensor nodes for further processing. Through this defini-tion, it is important to identify the following key elements : a physical sensor to collect the desired data, a protocol to communicate that data, an end-user application where data could be observed and analyzed. As is well known, WSN applications include, but are not limited to, surveillance and monitoring such as humidity, temperature, light, dust, pressure, motion detection, air pollution, fire detection, acoustic sensors, optical sensors, etc.
Internet of Things (IoT) is another important concept. IoT is a recent communication model in which everyday life objects are equipped with transceiver units, communication pro-tocols, and microcontrollers. These objects are able to communicate with the network and with one another, becoming a part of the Internet [Tan16]. As a matter of fact, wireless sen-sor network concept was an important building block for the IoT vision[Jac15]. In addition, urban IoTs enhanced the vision of smart cities [ZBC+14]. According to [CIS15], it is estima-ted that the number of devices, which will be connecestima-ted to the network, will reach about 50 billion devices by 2020. Other less pessimistic predictions estimate that the total number will be in the range of 20 - 30 billion connected things [KMF+14]. Regardless of the exact num-ber, a huge growth in the number of the connected devices is expected over the next few years. Let us take a closer look at the main blocks of a sensor node. A sensor node is composed of four main units [KAMH17] : sensing unit, processing unit, radio transceiver unit, and power unit as shown in figure 1.2. The sensing unit consists of a physical sensor which collects the desired data from the environment. The processing unit, which is composed of a processor and a specific operating system, processes data captured from nodes and transmits them to the network via the radio unit. The radio unit is responsible for all transmissions and receptions of data via defined communication protocols. The power unit supplies the energy to feed the different components.
Sensing Unit Processing Unit Radio Transceiver Unit
Power Unit
1.4 WSN protocols 19
Energy consumption, that is due to data sensing, processing, and communication, is one of the major challenges in wireless sensor networks. Sensor nodes have very limited energy resources, that imposes severe restrictions on processing capabilities, storage capacities, and transmit power. Consequently, the transmit power is very moderate to reduce the transmis-sion energy consumption since most of the energy usage is related to radio communications, which will, therefore, lead to short-range wireless communication links (depending on the communication protocol). This fact is exploitable in our context, as will be shown later on.
1.4
WSN protocols
One protocol cannot, indeed, cover all of the projected applications for WSNs and IoT. A large number of protocols are designed to satisfy a set of requirements based on the applica-tions. As shown in figure 1.3, WSN protocols can roughly be classified into three categories according to the following parameters : data rate, radio range, and power consumption (bat-tery lifetime). The first category is widely adopted to serve the applications that need to transmit moderate data rates over short ranges at low power consumption. The most well-known protocols for this category are the communication protocols that use the IEEE 802.15.4 standard (cf. subsection 1.4.1.1). The Second category offers a very long battery lifetime to serve the applications that need to send very limited amounts of data over long distances. This type of wireless networks is called a Low-Power Wide-Area Network (LPWAN). The most notable LPWAN technologies are SigFox (cf. subsection 1.4.2.1) and LoRA (cf. sub-section 1.4.2.2) radio standards. The third category uses the existing cellular technologies to serve the applications that need high data rates, but it is not widely used due to the high power consumption.It is important to underline that these categories should be considered separately when modeling radio propagation models, as will be shown in the next chapters.
Low transmit Power Short radio range
Short Range Communications Low Power Wide Area Cellular Networks
Long battery Life
Low transmit power Long radio range Long battery Life
High transmit power Long radio range Short battery Life
Average data rates Very low data rates High data rates
EDGE
1.4.1 Short range communication protocols
1.4.1.1 IEEE Std 802.15.4
IEEE 802.15.4 is a simple, low rate, low cost, short-range, flexible, and low power consump-tion standard for the lower layers of the OSI model. It defines the specificaconsump-tions of the physical layer (PHY) and also the medium access control (MAC) layer for a communication protocol which is designed to satisfy a set of simple needs for wireless communications with limited power and moderate throughput requirements. The standard was first defined in 2003 by the IEEE 802.15.4 working group, yet new variants of the original IEEE 802.15.4-2003 such as IEEE 802.15.4-2006, IEEE 802.15.4-2011, and IEEE 802.15.4-2015 were released afterward to add new features and clarifications to the existing versions.
The technical specifications are fully detailed in the document defining the IEEE 802.15.4 - 2003 standard [IEE03], but it is important to mention some of the characteristics of the IEEE 802.15.4 - 2003 that will be needed later on. The main characteristics of the standard can be summarized as follows :
• Layers : Physical and MAC layers.
• Data rates : 20 kb/s, 40 kb/s, and 250 kb/s. • Frequency bands :
– 868 MHz (1 channel) – 915 MHz (10 channels) – 2450 MHz (16 channels).
• Nominal transmit power of 0 dBm. • Receiver sensitivity :
– -92 dBm @ 868/915 MHz . – -85 dBm @ 2450 MHz.
• Modulation schemes : BPSK and O-QPSK.
• Low power consumption and short-range communications. • Peer-to-peer or star topology.
• Two different device types :
– Full-function device (FFD) : a node that can perform all the functions defined by the standard (i.e. send, receiver, and route data).
– Reduced-function device (RFD) : a node that can perform only a certain number of functions so they are considered as low power consumption devices since these nodes do not participate in traffic routing (i.e. end nodes).
1.4.1.2 IEEE 802.15.4 derived standards
IEEE 802.15.4 is the basis for a number of various higher layer standards such as ZigBee, WirelessHart, ISA100.11a, OCARI, 6LoWPAN, and MiWi. Although these standards share the same physical and MAC layers that have been defined in IEEE 802.15.4, they extend the protocol by developing the upper layers in the OSI model.
ZigBee : is one the most popular WSN protocols that is an IEEE 802.15.4-based protocol. The full ZigBee protocol stack defines the upper layers of the system all the way to the ap-plication layer.
1.4 WSN protocols 21
WirelessHart : is an open industrial communication technology based on the HART Com-munication protocol (Highway Addressable Remote Transducer). As its name indicates, it is as the wireless extension to the HART protocol. The protocol stack of WirelessHart imple-ments IEEE 802.15.4 physical layer at 2.4 GHz.
ISA100.11a : is a wireless communication technology developed by the International Society of Automation (ISA) to provide a secure and reliable wireless system for industrial applica-tions. Like WirelessHart, ISA100.11a is also based on IEEE 802.15.4 standard, implementing the 2.4 GHz physical layer. However, ISA100.11a standard modifies the IEEE 802.15.4 MAC layer, rather than adopting it in its entirety.
6LoWPAN : stands for IPv6 over Low-Power Wireless Personal Area Networks. 6LoWPAN has defined the mechanisms for the upper layers that allow IPv6 packets to be transferred over IEEE 802.15.4-based protocol.
MiWi : is an IEEE 802.15.4-based protocol designed by Microchip Technology, a well-known micro-controller manufacturer. The main feature of MiWi protocol is that it offers a smaller footprint, which makes microcontrollers perform opertaions with smaller memory.
OCARI : is a wireless communication protocol that was developed during the OCARI project that is funded by the French National Research Agency (ANR). The main goal of the project was to optimize the wireless communications for industrial networks. The project adopted the physical layer proposed a new MAC layer.
IEEE 802.15.4 derived standards
Standard ZigBee WirelessHart ISA100.11a OCARI 6LoWPAN
IEEE 802.15.4 2003 802.15.4 2006 802.15.4 2006 802.15.4 2006 802.15.4 2003 Frequency bands 868 MHz 915 MHz 2450 MHz 2450 MHz 2450 MHz 2450 MHz 868 MHz 915 MHz 2450 MHz Data rates 20 kbps 40 kbps 250 kbps 250 kbps 250 kbps 250 kbps 20 kbps 40 kbps 250 kbps Modulation BPSK
O-QPSK O-QPSK O-QPSK O-QPSK
BPSK O-QPSK Receiver sensitivity -92 dBm @915MHz -85 dBm @2450MHz -85 dBm -85 dBm -85 dBm -92 dBm @915MHz -85 dBm @2450MHz Nominal transmit power 0 dBm 0 dBm 0 dBm 0 dBm 0 dBm Radio range (roughly) 150 m 150 m 150 m 150 m 150 m
1.4.2 Low-Power Wide-Area Network (LPWAN) protocols
As was mentioned earlier, the number of connected nodes is expected to reach tens of billions of nodes over the few next years. However, some of these nodes need only very li-mited bandwidth to transmit small amounts of data over large distances. For many of these applications, the existing cellular systems are not well adapted to the specific needs of these applications because of the high power consumption, and the costly equipment. To this end, alternative communication systems are proposed to provide communications with the required objectives : long radio range, low data rates, and very low power consumption (i.e. a Low-Power Wide-Area Network (LPWAN)). SigFox and LoRA are the most well-known LPWAN technologies.
1.4.2.1 Sigfox
Sigfox is a proprietary protocol developed by the French company Sigfox [Sig]. It is a wide-range communication system that has been designed to offer a very low bitrate with significantly low power for remotely connected nodes using Ultra-Narrow Band (UNB) tech-nology. In fact, transmitting through UNB channels makes it possible to send data over large distances with low transmit power levels. The main features of the protocol are summarized as follows : • UNB technology (BW = 100 Hz). • Modulation scheme : – Uplink : D-BPSK. – Downlink :GFSK. • Bitrate :
– Uplink : 100 bps or 600 bps (depends on the operation region). – Downlink : 600 bps.
• Frequency bands (depends on the operation region and the local regulations) : – 868 - 868.2 MHz.
– 902 - 928 MHz. • Low radiated Power :
– 25mW = 14dBm @ 100bps. – 150mW = 22dBm @ 600bps. • Receiver sensitivity : – -142 dBm @ 100bps. – -134 dBm @ 600bps. • Link budget : ≈ 160 dB. • Uplink messages :
– Payload size for each message is from 0 to 12 bytes + 14 bytes of header. – Maximum 140 uplink messages per object per day to extend battery life. • Downlink messages :
– Payload size for each message is static : 8 bytes. – Maximum 4 downlink messages per object per day. • Low power consumption :
– 50 Microwatt with standby time of 20 years [PC15]. • Network System Architecture :
1.5 Simulation tools for wireless sensor networks 23
– Sigfox objects. – Sigfox base stations. – Sigfox cloud.
– Customer service (End-user). 1.4.2.2 LoRa
LoRa, an abbreviation for “Long Range”, is a proprietary comunication protocol for a LPWAN that is designed to achieve long-range and low power communication links. It is important to point out that LoRa is the physical layer for the protocol, while LoRaWAN defines the whole communication protocol and network architecture. LoRaWAN specifications vary depending on the operation region and on the local regulations. A brief summary of the LoRaWAN specifications [Lor15] can be given as follows :
LoRaWAN specifications
Europe North America
Frequency bands 863 - 870 MHz 902 - 928 MHz
Channel BW downlink 125 kHz 500kHz
Channel BW uplink 125/250kHz 125/500kHz
Modulation scheme FSK FSK**
Transmit power downlink 14 dBm 27 dBm
Transmit power uplink 14 dBm 20 dBm
Data rate 250 bps - 50 kbps 980 bps - 21.9 kbps
Link budget downlink 155 dB 157 dB
Link budget uplink 155 dB 154 dB
Maximum number of
messages per day Unlimited
Battery lifetime
(2000 mAh) 105 months
Network architecture
- LoRa Sensors
- Base Station (LoRaWAN Gateway) - Network Server
- End-user application Table 1.2 – LoRaWAN specifications
1.5
Simulation tools for wireless sensor networks
It is crucial for researchers and engineers to study, develop, test, and evaluate new de-ployments, protocols, algorithms, and applications. These performance tests can be conduc-ted through test-beds or simulation. Although test-beds are more realistic and more reliable as they conduct real experiments, they are fairly complex, time-consuming, costly, or even practically unfeasible for deploying a large number of nodes [AFAHN13]. Simulation is an ap-propriate alternative to examine, evaluate, and study network parameters before deployment especially for large-scale WSNs. In fact, simulations provide a cost-effective, fast-deployable, and fairly reliable solution. A comparison between simulation tools and testbeds is given
in [TWCM10] in order to provide a reference for choosing between them according to the research requirements. In fact, simulation is the research tool used by the majority of the research community as revealed by an investigation conducted through the papers published in SENSORCOMM conference [AAFA+12].
A WSN simulator allows constructing virtual and interactive networks in order to ob-serve and evaluate the operation of the network. A wide range of available WSN simula-tors is presented in the literature. Exhaustive surveys of WSN simulation tools are given in [NS15, Yur16, AAFA+12]. However, this subsection intends to give a short description of some widely used network simulators with a particular focus on the integrated radio propagation models, which have a major impact on the reliability and quality of the simulation results. The main idea, therefore, is to present the most widely used network simulators in order to identify the radio propagation models used in these simulators. It is important to note that those models will not be presented here, but rather they will be discussed in detail in chapter 2. There follows a brief summary of some well-known WSN simulators :
Network simulator 2 (NS-2)
NS-2 is one of the most popular WSN simulators [JZD09]. It is a discrete event open source network simulation tool which is written in object-oriented programming languages (i.e. C++ with OTcl). It is important to underline that NS-2 offers three simple non-realistic radio models [ns2a] [ns2b].
MannaSim
MannaSim is regarded as an extension for NS-2. This later extension takes the main features of NS-2 and adds ones to develop and analyze different WSN applications [Man]. MannaSim inherits the radio models from NS-2 [PRG15].
Network simulator 3 (NS-3)
NS-3 a is well-known, discrete-event, and open-source network simulator targeted mainly for educational and research purposes. NS-3 integrates a number of simple radio models for urban, suburban, and open environments [np16]. In fact, despite the fact that NS-3 offers a number of models that are widely used in the literature, these models are still not accurate enough or even not applicable in some circumstances.
TOSSIM
TOSSIM is a discrete-event simulator dedicated to networks running on the operating sys-tem TinyOS, which is an operating syssys-tem designed specifically for WSNs [TOSb]. TOSSIM offers simple radio propagation models with the possibility to add additive white Gaussian noise (AWGN) [TOSa].
Cooja
1.5 Simulation tools for wireless sensor networks 25
radio propagation model which abstracts the radio range simply as circles. It also includes a more sophisticated model i.e. a 2D site-specific radio propagation model [WGG10]. Although site-specific modeling provides accurate results, the computational complexity of this family of models is often not tolerable because simulation results should be obtained under time constraints.
Worldsens
Worldsens simulation environment is an integrated platform dedicated to designing ap-plications for sensor networks. The Worldsens environment consists of two simulation tools (WSim and WSNet) that are used independently or together during the application design [FCF07] :
— WSNet is an event-driven large-scale wireless sensor network simulator. WSNet offers a number of basic radio propagation models : disk model (range), simple radio pro-pagation models, or even probability distribution (statistical characterizations of the radio channel) [WSN].
— WSim is a hardware platform simulator that uses microcontroller binary codes. QualNet
QualNet a commercial simulation platform developed by Scalable Network Technologies, Inc. Although it is not dedicated to WSN simulations, it supports WSN simulations using 802.15.4 library. The manual of the model library index of QualNet 8.0 [Inc17] shows that the urban propagation library includes a number of simple propagation models.
OMNeT++
OMNeT++ is an extensible component-based platform for building network simulators [omn]. Independent plugins are developed to provide specific functionality for the platform such support for WSNs :
— Castalia is a plugin for OMNeT++ platform developed by the National ICT Australia. It was designed for simulating WSNs and body area networks (BANs). Castalia users manual [Cas13] gives an overview of the radio models supported by this simulator. It shows that Castalia integrates simple radio propagation models. In addition, it also offers temporal models to describe the temporal variation of the channel.
— MiXiM is a plugin for OMNeT++ platform designed for WSNs, BAN, vehicular net-works, etc. MiXiM includes the simple radio propagation models that are widely used by the presented network simulators [KSW+08, MiX].
OPNET
OPNET Modeler Wireless Suite supports a wide range of wireless networks such as cellular networks, ad-hoc networks, wireless LAN (IEEE 802.11), WSNs, and satellite communications [Beu12]. Like other presented simulators, OPNET provides also several simple propagation models [HBVP17]. It is necessary to underline that OPNET is now part of Riverbed company,
the new name of OPNET Modeler Suite is Riverbed Modeler. CNET
CNET is an open source network simulator developed for research purposes at the Uni-versity of Western Australia. It implements the IEEE 802.11 standard, which makes it able to simulate wireless networks. It integrates a simple radio propagation model, but it allows the users to define their own radio propagation models [CNE].
Since our main aim was, as mentioned earlier, not to make an exhaustive list of all the WSN simulators, but rather to give an overview of the radio models implemented in that network simulators. In fact, the review leads to the conclusion that well-known network si-mulators are implementing either simple radio propagation models which are environment independent and often not accurate enough ; or sophisticated site-specific radio propagation models which can give more accurate simulation results, but on the other hand, these simu-lators do not address time constraints. Hence, a new effective way for radio channel modeling for network simulators is required, taking into account the addressed problem.
1.6
PERSEPTEUR project
As mentioned in the introduction, this thesis is a part of a national project called “PER-SEPTEUR”. The project’s name is an abbreviation for the complete French name of the project ”PlateformE viRtuelle 3D pour la Simulation des rEseaux de caPTEURs” which can be translated as a 3D Virtual Platform for Wireless Sensor Network Simulation. PERSEP-TEUR is a research project funded by the French National Research Agency (ANR) and supported by the National Center for Scientific Research (CNRS).
1.6.1 Project objectives
The primary objective of this project is to construct realistic simulation platforms for WSNs and IoT for outdoor smart city applications using free 3D city models. The simulation tools of this project must satisfy the following global requirements :
— Provide fairly accurate simulation results within reasonable computational time. — Take into account the geometrical details of the simulation environment.
— Able to represent mobile nodes and dynamic environments. — Able to support large-scale wireless sensor networks.
The final simulation platform should include the following features :
• Simulation and visualization of WSNs services in a 2D/3D environment.
• Improving the quality of sensor deployment in terms of communication links feasibility and reliability by :
– Providing accurate radio propagation predictions.
– Detecting potential interference areas among the network in order to improve the link quality by placing sensor nodes in an optimal way.
1.6 PERSEPTEUR project 27
– Considering real environments.
• Visualization of the simulation results in public GIS platforms such as OpenStreetMap. • Considering node mobility (i.e. public transport).
• Providing support for academic staff to illustrate the concept of WSNs using 3D virtual environments.
1.6.2 Project parteners
PERSEPTEUR project involves academic and corporate partners. More precisely, it is a collaboration between three research laboratories and a R&D engineering company spe-cializing in 3D virtual representations. Each partner will provide an essential building block for constructing the final platform. To this end, the project is structured around four major axes that correspond to four different tasks. It should be pointed out that the aforementioned global project requirements will be reflected in each task. For the sake of brevity, the project partners will be listed with a brief description of the main role of each partner :
XLIM research institute is responsible for modeling the radio channel between the nodes. Channel modeling must surely take into account the global project specifications presented in 1.6.1, which are regarded technically as a set of constraints.
IEMN research institute is responsible for developing interference models in order to eva-luate the impact of a set of nodes on a given link.
Lab-STICC is responsible for providing the main kernel of the project, which is a simulation tool for sensor networks (called CupCarbon). CupCarbon is a multi-agent and discrete-event WSN and IoT simulator for both educational and research purposes [MLBK14].
Virtualys is responsible for providing a 3D model of the city of Brest in the form of a virtual 3D environment.
The architecture of simulation platform consists of many modules as shown in figure 1.4 : — City model module : it represents the city in digital form (2D or 3D). It includes
information about the simulation scene (buildings, objects, roads, bridges, etc.). — Mobility module : it defines the routes and trajectories of the mobile nodes. — Network module : it defines the scenario and the parameters of the WSN to be
simulated.
— Communication Script module : this module interprets the SenScript language (i.e. the script used to program the sensor nodes in CupCarbon simulator) in order to allow the simulator to understand and execute the instructions of the script.
— Radio Propagation module : it models the radio channel between the communi-cating nodes in the network. The radio channel between each pair of nodes could be represented in matrix form (i.e. propagation and connection matrix), which estimates the quality of radio links between the communicating nodes.
— Interference module : this module is models the effect of the network nodes on a given link used to determine if the sent message could be received by the receiver. — Simulation module : this module is based on a discrete event simulation.
Figure 1.4 –Simulator structure.
1.7
Thesis motivation and objectives
Up to now, the first two parts have tended to present the context of the thesis. Henceforth, the discussion will focus on the research question of the here presented thesis.
As mention earlier, simulations provide a cheaper and more practical way to test and ex-plore different deployment scenarios. For researchers and engineers, simulation is an essential tool to verify the behavior and performance of the designed network before moving towards live implementation. On the other hand, simulation results could give deceptive results. In fact, simulation results obtained by WSN simulators are affected by several factors, one of which is by modeling the radio channel. Although the simulation results are considerably af-fected by the radio channel characterization, well-know WSN simulation tools (cf. section 1.5) use simple and non-realistic radio propagation models, which can lead to erroneous results. Consequently, it is inevitable to integrate more accurate radio models into simulation tools in order to get more realistic results.
The quality of radio simulation results strongly depends on the degree of realism of mo-deling the propagation environment and also on the numerical methods that are used to characterize the propagation mechanisms. On the one hand, increasing the number of mo-deled elements will considerably increase the computational complexity and will occupy the available physical resources. This could be prohibitive due to the lack of available resources (physical constraints or time constraints).
It is difficult or even infeasible, under strict time constraints, to give realistic performance, to deal with realistic propagation environment, to execute numerical methods, and to treat a large number of nodes some of which are mobile. To that end, it is necessary to simplify the simulation scenario by excluding certain level of details (propagation scene, propagation mechanisms) in order to reduce the computational complexity. This can be done by redu-cing the complexity of the propagation scene, using less greedy numerical methods, adopting acceleration techniques, or sacrificing less significant propagation mechanisms but within a
1.7 Thesis motivation and objectives 29
manageable margin of error. Nevertheless, this brings up the question of how to define a trade-off between accuracy, complexity, and execution time for the radio propagation models. Let us formulate the research problem, the thesis intends to propose radio propagation models that can accurately and quickly predict radio propagation behavior for wireless sensor networks and smart city applications. To this end, a set of trade-offs between accuracy and computational time will be defined to meet the following requirements :
– Models that are able to provide accurate results under severe time constraints.
– Models that consider the propagation environment –> that requires site-specific models. – Models that are able to deal with large-scale networks (i.e. up to thousands of nodes)
–> that requires very fast models.
– Models that are able to support mobile nodes –> that requires quasi-instantaneous mo-dels.
In the literature, a very large number of radio models has been proposed to predict radio waves behavior. Radio propagation models are broadly classified into two families [GMM+17] : Empirical models
Empirical propagation models are a set of equations extracted for a particular environ-ment and in particular conditions (i.e. antenna heights, frequency range, indoor/outdoor, urban/suburban/rural, etc.). These models are derived from extensive and time-consuming field measurements. Despite the fact that the empirical propagation models can be regarded as environment-dependent but non-site-specific, they do not require detailed information about the propagation environment but rather some indicative information. Empirical models are attractive because they are extremely fast and easily implementable. On the other hand, they fail to provide accurate results.
Deterministic models
Unlike empirical propagation models, deterministic propagation models do not require field measurements. On the other hand, deterministic modeling requires detailed information about the propagation environment [KM16]. This class of models depends on exact or approxi-mate numerical solution of Maxwell’s equations. Therefore, they require high computational time.
According to project requirements, site-specific (i.e. deterministic) radio propagation mo-dels are required. It is important to underline that the computational complexity for site-specific models is a critical issue especially for large-scale simulations because this class of models needs a huge amount of data related to the information of the propagation envi-ronment. In addition, It solves complex numerical methods. Hence, classical deterministic propagation models cannot be directly adopted in the research problem. On the other hand, the empirical class is extremely fast and can easily be implemented but it does not provide accurate estimations in such environments, which leads to exclude also this class of models. In fact, it is difficult to guarantee an ultimate degree of precision under such severe time constraints. Hence, the core issue of the here presented thesis is to place the cursor at the
most suitable trade-offs that give the best balance between accuracy and computational time as shown in figure 1.5.
Less accurate
More accurate
Fast, easily implementable
Computationally complex
Empirical models
Trade-off
Deterministic models
Figure 1.5 –Accuracy computational time trade-off.
1.8
Thesis Contributions
This dissertation proposes efficient, fast and accurate numerical methods for compu-ting/modeling electromagnetic waves in realistic environments. The proposed models are adapted to meet the expected project specifications to the maximum extent possible by ta-king advantage of the specificity of the radio channels in this context. The final models are deterministic models that adopt a set of effective acceleration techniques in order to reduce the computational complexity with a minimal loss of accuracy. These models will be integra-ted into a WSN simulation platform in order to provide a more realistic and more reliable platform.
1.9
Conclusion
In this chapter, the smart city concept was introduced. Smart cities offer several important applications and diverse interesting uses that facilitate people’s everyday life. Wireless sensor networks are an important element in smart cities. Many communication protocols for WSNs were proposed to cover a wide range of requirements. The simulation of sensor networks is an indispensable tool for testing and validating before real deployment. The simulation tools are also important to the researchers in order to test and validate their protocols and al-gorithms. In fact, real experiments are costly, time-consuming and difficult, especially when talking about a large number of nodes distributed over large areas. For this reason, network simulators should be reliable, accurate, and fast.
PERSEPTEUR project was presented in this chapter. Its objective is to construct a rea-listic and accurate simulator. An important element of the project’s requirements is the radio propagation models since they affect considerably the simulation results. In fact, it was shown that existing well-known WSN simulators offer simple radio propagation models. Moreover, existing radio propagation models cannot be directly adapted to the simulator required by the project because they do not satisfy the project requirements in terms of accuracy and time constraints. A set of trade-offs will be proposed to address this problem.
Chapter 2
Background and state of the art
Contents
2.1 Introduction . . . 33
2.2 Radio Channel fundamentals . . . 33
2.2.1 Multipath propagation . . . 33
2.2.2 Different Scales of Attenuation . . . 36
2.3 Parameters of Multipath Radio Channels . . . 37
2.3.1 Time dispersion parameters . . . 37
2.3.2 Coherence bandwidth . . . 38
2.3.3 Doppler spread and coherence time . . . 38
2.3.4 Narrowband and wideband channels . . . 39
2.3.5 Macro, Micro, Pico, and Femto Cells . . . 40
2.4 Review of empirical propagation radio models for WSNs . . . 41
2.4.1 Free Space Model . . . 41
2.4.2 Adapted Free Space Model . . . 42
2.4.3 Simplified Two-Ray Ground Reflection Model . . . 42
2.4.4 Free Space and Simplified Two-Ray Hybrid Model . . . 44
2.4.5 Two-Ray Ground Reflection Model . . . 44
2.4.6 Free Outdoor Model (FOM) . . . 45
2.4.7 Log Normal Model Shadowing Model . . . 46
2.4.8 Survey of radio propagation models in WSN Simulators . . . 48
2.5 Deterministic modeling . . . 50
2.5.1 Maxwell’s Equations . . . 50
2.5.2 Electrical properties of propagation medium . . . 51
2.5.3 Rigorous solutions to Maxwell’s equations . . . 52
2.5.4 Helmholtz wave Equation . . . 52
2.5.5 High frequency asymptotic methods . . . 52
2.6 Ray Tracing Techniques for Radio Propagation . . . 61
2.6.1 Ray-launching method . . . 61
2.6.2 Image method . . . 62
2.6.3 Hybrid method . . . 63
2.6.4 Other methods . . . 63
2.7 Ray-Tracing Acceleration Techniques . . . 63
2.7.1 Dimension Reduction . . . 63
2.7.3 Visibility Graphs . . . 64
2.7.4 GPU Acceleration . . . 64
2.8 Formulation of the research problem . . . 64
2.9 Proposed Solution . . . 64
2.1 Introduction 33
2.1
Introduction
Chapter 1 outlined the context, motivation and expected outcomes of the thesis. Chapter 2 aims to provide a synopsis of the relevant literature in order to suggest the most efficient proposition for the research problem. To that end, chapter 2 will cover several aspects such as radio channel fundamentals, a review of empirical channel modeling for WSNs, deterministic modeling principles. The chapter concludes by formulating the research problem and then by explaining the adopted proposal.
2.2
Radio Channel fundamentals
Radio propagation channel is the medium where the electromagnetic waves propagate bet-ween the transmitter and the receiver. It is an inevitable element for any radio communication system. Through the radio channel, information is transmitted over a given bandwidth by es-tablishing point-to-point or point-to-zone radio links. However, real radio channels introduce a variety of impairments and disturbances to the signal. Understanding the radio channel is thus a key element in the implementation and design of wireless communication systems. It is necessary here to clarify what is the difference between the radio channel and the transmission channel. A transmission channel refers to the radio channel but it includes also the antenna radiation patterns of the transmitter and receiver.
For any radio communication system, radio channel characterization is a crucial issue for system design, because the quality of radio links is the main limiting factor of the end-to-end communication quality. One possible way to obtain exact information about the radio channel is to perform field measurements. However, field measurements for complex test scenarios could be prohibitively expensive and very time-consuming because all transmitters, receivers, and all other equipment must be deployed over a large area, including possibly a relatively large number of test locations. It is, therefore, essential to develop effective and accurate radio propagation models as a viable alternative to field measurements.
2.2.1 Multipath propagation
Radio propagation is the behavior of radio waves as they move from one point to ano-ther in the propagation environment. When the radio waves travel from the transmitter to the receiver, they encounter many obstacles in their propagation way, especially in urban environ-ments. Knowing the phenomena that influence radio waves is indispensable for characterizing the behavior of the radio channel. The principal phenomena can generally be described by four basic mechanisms or interactions : reflection, diffraction, refraction (also called transmission), and scattering. At the receiver side, the signal will be received via a set of different paths undergone multiple interactions, as depicted in figure 2.1. These interactions can affect the direction, amplitude, phase, and polarization of the initial radio wave. Consequently, the re-ceived signal is composed of a number of attenuated, phase shifted, and time-delayed replicas of the transmitted signal. This propagation mechanism is known as multipath propagation, which can guarantee a good-quality radio communication over non line of sight (NLOS) confi-gurations.
Reflection occurs when radio waves encounter relatively large objects (i.e. greater than the wavelength such as buildings, walls, grounds, etc.). When a radio wave travels through one medium and hits another medium having electrical properties different from that of the first medium, part of the energy is reflected into the initial medium and another part is transmitted into the second medium. In outdoor environments, It is generally assumed that the transmitted field is totally absorbed by the building. It is therefore considered as non-significant propagation phenomenon in outdoor scenarios.
Diffraction occurs when radio waves encounter edges, corners, or sharp obstacles. Diffrac-tion allows signal propagaDiffrac-tion behind obstacles in the shadowed regions. Signal field strength becomes more attenuated, as the receiver goes deeper behind the shadowed (obstructed) re-gion. Diffraction can be understood by the Huygens’ principle [SS99] that says that each point on a wavefront is regarded as a point source to produce secondary spherical wavelets and the sum of those wavelets forms another spherical wavefront at the same direction.
Scattering occurs when radio waves encounter rough surfaces, small objects, or atmosphere particles [HHC06], so that the reflected wave is diffused in all directions. Trees, signpost, lamp-post, and other similar small objects tend to scatter the incident waves in all directions.
B Reflection Reflection Diffraction Scattering Tx Rx Diffraction
Figure 2.1 –Urban propagation scenario.
Mathematical model for multipath radio channel :
It is fundamental to characterize the wideband multipath channels by mathematical mo-dels, for many purposes. The key purpose is to understand how the channel behaves and how it does affect the transmitted signals. An additional crucial purpose is to use these models for computer simulations. Finally, it is possible through these models to quantify the radio chan-nel by some parameters such as the rms delay spread, coherence time, coherence bandwidth, and Doppler spread (cf. section 2.3). These parameters are useful in comparing different mul-tipath radio channels and in developing broad guidelines for wireless networks designers.
As indicated previously, the received signal in multipath environments is composed of a set of attenuated, phase shifted, and time-delayed replicas of the transmitted signal. Hence, it is convenient to view the radio propagation channel as a filter, which can be characterized
2.2 Radio Channel fundamentals 35
by its impulse response [Ham13]. The complex impulse response h(τ ) of a radio channel can be modeled (assuming that the channel is time-independent over the interval of interest ) by a time-invariant linear filter and is expressed as :
h(τ ) =
NT X
n=1
ane−jθnδ(τ − τn) (2.1)
where NT is the total number of paths between a given transmitter/receiver pair, an is the
amplitude of the nth path, τnis the delay of the nth path (depends on its length), and finally
θn is the phase of the nth path (depends on its length, on the frequency, and on the
electro-magnetic interactions that occurred during the propagation).
Spatial variability : is due to the mobility of the receiver and/or transmitter in the propa-gation environment. In typical multipath environments, the received signal fluctuates rapidly over small distances, because as the receiver and/or the transmitter moves, the received paths are not the same anymore. Hence, the received signal depends on the properties of the new/modified paths at the receiving side.
Time variability : is due in part to the mobility in the propagation environment, modifying the characteristics of the paths. New paths may appear, other existing paths may disappear, even the persevered paths are subject to change. The transmitter and receiver are assumed to be fixed.
Thus, the channel time-variant impulse response h(t, τ ) is modeled by a time-variant linear filter and is expressed as :
h(t, τ ) =
NT(t) X
n=1
an(t)e−jθn(t)δ(τ − τn(t)) (2.2)
where NT(t), an(t), τn(t), and θn(t) are defined as before but they are time-dependent
para-meters.
Although the radio channel can be characterized by the complex impulse response h(t, τ ), it can also be modeled by other system functions, which are called Bello’s functions. These functions are related through Fourier transforms as shown in figure 2.2 :
— Delay spread function h(t, τ ) : is the time-variant impulse response, which is called by Bello the input delay spread function.
— Time-variant transfer function H(t, f ) : is simply the Fourier transform of the delay spread function h(t, τ ) with respect to the time delay τ . It is interpreted as the time evolution of the transfer function, given by
H(t, f ) = Z ∞
−∞
h(t, τ ) · e(−j2πf τ )dτ (2.3) — Delay Doppler spread function S(τ, v) : represents the spreading of the input
signal in the delay and Doppler domains, given by S(τ, v) =
Z ∞
−∞
— Doppler spread function D(f, v) : is the Fourier transform of the delay Doppler spread function with respect to τ , that is
D(f, v) = Z ∞
−∞
S(τ, v) · e(−j2πf τ )dτ (2.5)
Delay spread function
Doppler-spread function Time-variant transfer Function Delay Doppler-spread Function ) , ( t h ) , ( tf H F 1 f F t F 1 v F ) , ( vf D ) , ( v S 1 v F F 1 f F t F
Figure 2.2 –Bello functions.
where τ is the time delay, t denotes the time, v designates the Doppler shift, and f is the carrier frequency. A detailed description of Bello functions is given in [Bel63].
2.2.2 Different Scales of Attenuation
In[Lee10], Lee has shown that the received signal strength can be expressed as the product of two terms :
r(t) = m(t) · r0(t) (2.6)
where r(t) is the received signal strength, m(t) characterizes the large-scale fading, and r0(t)
characterizes the small-scale fading.
Large-scale fading : describes the average attenuation of signal power over relatively large distances (i.e. it gives the envelope of the received signal over large distances). Large-scale fading can be further classified according to its cause into path loss and shadowing. Path loss is the mean attenuation that depends mainly on the frequency and on the distance between the transmission point and the reception point, while shadowing occurs when the signal is obs-tructed by an obstacle in the propagation environment, which can attenuate the signal power. Small-scale fading describes the fast fluctuations of the signal power around its local mean value due to the constructive or destructive sum of multipath components. These fluctuations occur over short time intervals and/or over short distances because of the mobility of the communicating terminals, and/or because of the movement of the objects in the propaga-tion environment. Small-scale fading is typically modeled by statistical distribupropaga-tions such as
![Figure 1.1 – Smart city applications. Image source : Trends in smart city development : Case studies and recommendations [DS16].](https://thumb-eu.123doks.com/thumbv2/123doknet/14506785.720246/22.918.160.753.476.874/figure-smart-applications-image-trends-development-studies-recommendations.webp)
